Detecting Malware with an Ensemble Method Based on Deep Neural Network

Table 2

The list of two CNN network structures parameters.

	Network layer type	Size	Output dimension
BaseNet	Input layer	-	(1, 1, 64, 64)
	Convolutional Layer	32 5 × 5 Convolution kernel	(1, 32, 64, 64)
	Max pooling layer	2 × 2, stride 1	(1, 32, 32, 32)
	Dropout layer	-	(1, 32, 32, 32)
	Convolutional layer	64 5 × 5 Convolution kernel	(1, 64, 32, 32)
	Max pooling layer	2 × 2, stride 1	(1, 64, 16, 16)
	Dropout layer	-	(1, 64, 16, 16)
	Fully connected layer	Logistic regression	(1024, 1)
	Dropout layer	-	(1024, 1)
	Output layer	-	1
VGGNet	Input layer	-	(1, 1, 64, 64)
	Convolutional layer	32 3 × 3 Convolution kernel	(1, 32, 64, 64)
	Convolutional layer	16 3 × 3 Convolution kernel	(1, 16, 64, 64)
	Max pooling layer	2 × 2, stride 1	(1, 16, 32, 32)
	Dropout layer	-	(1, 16, 32, 32)
	Convolutional layer	32 3 × 3 Convolution kernel	(1, 32, 32, 32)
	Max pooling layer	2 × 2, stride 1	(1, 32, 16, 16)
	Dropout layer	-	(1, 32, 16, 16)
	Fully connected layer	512 maxout unit	(32, 512)
	Dropout layer	-	(32, 512)
	Fully connected layer	Logistic regression	(32, 1)
	Dropout layer	-	(32, 1)
	Output layer	-	1